Benzo[b]quinolizinium Derivatives Have a Strong Antimalarial Activity and Inhibit Indoleamine Dioxygenase

Abstract

The heme-containing enzymes indoleamine 2,3-dioxygenase-1 (IDO-1) and IDO-2 catalyze the conversion of the essential amino acid tryptophan into kynurenine. Metabolites of the kynurenine pathway and IDO itself are involved in immunity and the pathology of several diseases, having either immunoregulatory or antimicrobial effects. IDO-1 plays a central role in the pathogenesis of cerebral malaria, which is the most severe and often fatal neurological complication of infection with Plasmodium falciparum. Mouse models are usually used to study the underlying pathophysiology. In this study, we screened a natural compound library against mouse IDO-1 and identified 8-aminobenzo[b]quinolizinium (compound 2c) to be an inhibitor of IDO-1 with potency at nanomolar concentrations (50% inhibitory concentration, 164 nM). Twenty-one structurally modified derivatives of compound 2c were synthesized for structure-activity relationship analyses. The compounds were found to be selective for IDO-1 over IDO-2. We therefore compared the roles of prominent amino acids in the catalytic mechanisms of the two isoenzymes via homology modeling, site-directed mutagenesis, and kinetic analyses. Notably, methionine 385 of IDO-2 was identified to interfere with the entrance of l-tryptophan to the active site of the enzyme, which explains the selectivity of the inhibitors. Most interestingly, several benzo[b]quinolizinium derivatives (6 compounds with 50% effective concentration values between 2.1 and 6.7 nM) were found to be highly effective against P. falciparum 3D7 blood stages in cell culture with a mechanism independent of IDO-1 inhibition. We believe that the class of compounds presented here has unique characteristics; it combines the inhibition of mammalian IDO-1 with strong antiparasitic activity, two features that offer potential for drug development.

INTRODUCTION

Tryptophan metabolism along the kynurenine (Kyn) pathway is a central component in neurodegenerative disorders and neuronal damage. The main metabolites of the kynurenine pathway include kynurenic acid, an antagonist of glutamate and nicotinic receptors; quinolinic acid, an agonist of N-methyl-d-aspartate receptors; and picolinic acid. Due to its potential role in diseases such as cancer (1), Alzheimer's disease (2), depression (3), stroke (4), and HIV infections (5), the kynurenine pathway is intensely studied with regard to chemotherapeutic applications in humans. Moreover, the pathway has been implicated in parasitic infections, such as trypanosomiasis (6), toxoplasmosis (7, 8), and cerebral malaria (CM), as shown in mouse model systems (9, 10). CM is a potentially fatal consequence of infection with Plasmodium falciparum and can lead, in case of survival, to neurological or cognitive deficits (10). Severe malaria remains a global health problem. Although the mortality rate decreased by about 25% between 2000 and 2010, malaria caused 627,000 deaths in 2012. Currently, no specific treatment for CM is available, but a couple of immunomodulatory therapies are under investigation (11).

As an initial step in the kynurenine pathway, tryptophan can be oxidized by heme-dependent indoleamine 2,3-dioxygenase 1 (IDO-1) and IDO-2 or tryptophan 2,3-dioxygenase (TDO), resulting in the formation of N-formylkynurenine (N-FKyn) and, subsequently, kynurenine. TDO is mainly expressed in the liver and is highly specific for l-tryptophan (l-Trp), while IDO-1 is ubiquitously expressed and has a broader substrate specificity for tryptophan compounds (12). Human IDO-1 (hIDO-1) and mouse IDO-1 (mIDO-1) have a sequence identity of 62% and high degrees of structural and functional similarity. IDO-1 and -2 of both mice and humans are encoded by tandemly arranged genes, indicating that the two isoforms arose via gene duplication. The sequences of human and mouse IDO-2 show a similarity of approximately 43% to human and mouse IDO-1, respectively, which results in distinct substrate ranges and kinetic activities, different selectivities toward some inhibitors, and distinct expression patterns with regard to tissues and stimuli, which make a complete functional redundancy of IDO-1 and IDO-2 unlikely (13–16). While IDO-1 is encoded only by the genomes of mammals, IDO-2 can also be found in lower vertebrates (14).

IDO-1 is not constitutively expressed but requires stimulation by host- and pathogen-derived inflammatory molecules such as proinflammatory cytokines (gamma interferon [IFN-γ], tumor necrosis factor [TNF]) and Toll-like receptor ligands (17), a regulation that functions as part of an immune response to pathological conditions such as protozoan infections (9, 18, 19). High levels of expression of IDO-1 lead to Trp depletion and increased levels of downstream neuroactive kynurenine metabolites (20). During systemic inflammation, including malaria infection, IDO is involved in vascular relaxation and blood pressure regulation (21). Moreover, tryptophan degradation and local deprivation by IDO function as counterregulatory pathways that mediate T-cell responses (17). IDO-2 is important for tumor escape and survival and is involved in immune tolerance (22).

A number of studies have indicated that IDO is involved in CM pathogenesis, as studied in experiments with human patients (23, 24) and mouse model systems (25–28). IDO-1 expression and activity are increased in the brains of mice during CM (25, 27), and IFN-γ, an important regulator of IDO-1, is a central element in the pathogenesis of CM (25, 29), although mice genetically deficient in IDO-1 do not seem to be protected against CM (9). Similarly, Mycobacterium tuberculosis infection induces IDO-1 expression and, thus, tryptophan metabolism, but a lack of IDO-1 activity did not affect survival after M. tuberculosis infection (30). Although its expression is not elevated during CM (13), IDO-2 might be able to compensate for a lack of IDO-1 activity. Nevertheless, CM is clearly correlated with altered levels of neuroactive kynurenine metabolites, including kynurenic acid, quinolinic acid, and picolinic acid (9, 23–25).

The potential of IDO inhibition for the treatment of CM and other diseases, including cancer, motivated the development of a range of IDO-1 inhibitors (1, 25, 29). When the therapeutic inhibition of IDO-1 is considered, it is important to mention that IDO-1 can have contradictory functions: it can exert immunoregulatory (Leishmania major) and antimicrobial (Toxoplasma gondii, Trypanosoma cruzi) effects, which are specific for the individual pathogen (31), with the molecular mechanisms being largely unknown. The most frequently used IDO inhibitor is 1-methyl-d-tryptophan (1-MDT), which has reached phase 1 clinical trials as an adjunct to conventional chemotherapy of cancer, since many tumors express IDO for immune suppression (15).

It has been shown that cationic berberine-type alkaloids exhibit, among other biological activities, antimalarial properties (32). Along these lines, several derivatives of berberine have been investigated to assess structure-activity relationships, and it has been shown that among them the parent berberine (compound 1a), its isomer (compound 1b), as well as C-13-alkylated derivatives, such as compound 1c, have the highest in vitro activity against Plasmodium falciparum (33, 34). In this context, the class of annelated quinolizinium derivatives may offer a promising starting point for the development of novel lead structures for antimalarial drugs, because the structure of these compounds closely resembles the cationic hetarene structure of berberine alkaloids and because their substitution pattern may be varied in a very broad way (35). Although annelated quinolizinium derivatives are well-established DNA binders (36), the interaction of these compounds with proteins and enzymes has hardly been investigated (34). Hence, to assess the propensity of quinolizinium derivatives to act as antimalarial agents, we compiled a compound library of known quinolizinium derivatives, compounds 2 to 5 (Fig. 1), and analyzed (i) their potential with respect to inhibition of IDO-1 and IDO-2, (ii) their in silico binding mode, (iii) their cytotoxicity for two human tumor cell lines, and (iv) their effect on P. falciparum in vitro and P. berghei in vivo. Moreover, we studied the active site of IDO-2 in comparison to that of IDO-1 via homology modeling, site-directed mutagenesis, and kinetic analyses.

FIG 1.

Structures of annelated quinolizinium derivatives (the IUPAC numbering is shown at compounds 3a to 3h). OMe, methoxy; n-Bu, n-butyl; Me, methyl.

MATERIALS AND METHODS

Site-directed mutagenesis of mIDO-2.

The cloning of mIDO-1 (GenBank accession number NM_008324) and mIDO-2 (GenBank accession number BC026393) was described previously (14, 15). On the basis of a detailed analysis of the heme environment in mIDO-2 in comparison to the crystal structure of hIDO-1 (described below), three mutants of mIDO-2 were generated via site-directed mutagenesis: mIDO-2 with an M-to-L change at position 385 (mIDO-2^M385L), mIDO-2 with an Y-to-F change at position 231 (mIDO-2^Y231F), and mIDO-2 with an A-to-D change at position 384 (mIDO-2^A384D). Mutations were introduced into mIDO-2 by PCR with Pfu DNA polymerase (Promega, USA) using mutated primers (for mIDO-2^Y231F, sense primer 5′-CCCAGACATATTTTTCTCGGTCATCC-3′ and antisense primer 5′-GGATGACCGAGAAAAATATGTCTGGG-3′; for mIDO-2^A384D, sense primer 5′-GGGTACTGACATGCTGAGCTTCTTG-3′ and antisense primer 5′-CAAGAAGCTCAGCATGTCAGTACCC-3′; for mIDO-2^M385L, sense primer 5′-GTACTGCCCTGCTGAGCTTCTTGAA-3′ and antisense primer 5′-TTCAAGAAGCTCAGCAGGGCAGTAC-3′; the mutated nucleotide is underlined). Methylated nonmutated template plasmids were digested with DpnI, and competent Escherichia coli XL1-Blue cells (Stratagene) were subsequently transformed. After the correct mutations were confirmed by sequencing, the genes were cloned into pDEST17.

Heterologous overexpression of mIDO-1 and -2.

E. coli Rosetta(DE3) cells containing pDEST17 with mIDO-1 were cultivated in Luria-Bertani (LB) medium containing 50 μg/ml carbenicillin and 35 μg/ml chloramphenicol as described previously with minor modifications (15). Rosetta(DE3) cells were grown in LB medium with antibiotics at room temperature until an optical density at 600 nm (OD₆₀₀) of 0.3 to 0.4 was reached. To increase the heme content in IDO, the medium was supplemented with 5-aminolevulinate (ALA), a natural precursor of heme. To induce overexpression, 25 μM IPTG (isopropyl-β-d-thiogalactopyranoside) and 0.5 mM ALA-HCl were added, and the cells were grown at room temperature overnight. Cells were harvested by centrifugation at 10,000 × g for 15 min at 4°C. The pellet was resuspended in 25 mM Tris, pH 7.0, and 150 mM NaCl in the presence of 10 μM phenylmethylsulfonyl fluoride, 150 nM pepstatin A, and 40 nM cystatin and stored at −20°C.

The mIDO-2 wild type (WT) and mutants were expressed in E. coli KRX as described previously (15). KRX cells were grown in Terrific Broth (TB) medium containing 50 μg/ml carbenicillin at 37°C to an OD₆₀₀ of 0.8. The culture was transferred to room temperature and further cultivated until the OD₆₀₀ reached 1.0 and 1.5. Expression was induced with 0.1% (wt/vol) rhamnose and 0.5 mM ALA-HCl, and the cells were grown overnight. The cells were harvested via centrifugation and resuspended as described above for mIDO-1.

Purification of mIDO-2 variants.

Cells were lysed in the presence of lysozyme and DNase for 60 min on ice, sonicated, and centrifuged (38,000 × g for 30 min at 4°C). The supernatant was applied to an Ni-nitrilotriacetic acid agarose column (Qiagen). After removing nonspecifically bound proteins, the His-tagged mIDO variants were eluted with 200 to 500 mM imidazole. To increase the stability of mIDO-2, 20% (vol/vol) glycerol was added to the eluted protein. The purity of the protein samples was controlled with SDS-PAGE, and protein concentrations were determined according to the Bradford method with bovine serum albumin as a standard (37). Protein immunoblotting was performed by using semidry Western blots with a monoclonal anti-His₆ antibody (Dianova in 5% nonfat milk with Tris-buffered saline and Tween 20 [TBST]) and a phosphatase-conjugated anti-mouse immunoglobulin antibody (Dianova in 5% nonfat milk with TBST) as the secondary antibody.

Determination of kinetic parameters of rmIDO-2 proteins.

IDO activity was determined by employing a methylene blue (MB)-ascorbic acid (AA) artificial reducing system in a 96-well microtiter plate as described previously (15, 38) with minor modifications. The assay was performed using freshly purified protein, whereby the final glycerol concentration was ≤1% (vol/vol). The standard reaction mixture (200 μl) contained 100 mM potassium phosphate buffer, pH 7.4, 100 mM KCl, 20 mM AA (neutralized with NaOH), 200 μg/ml catalase, 10 μM methylene blue, various concentrations of l-Trp, and 75 or 125 μg/ml mIDO-2. After 60 min of incubation at 37°C, the reaction was stopped by adding 40 μl of 30% (wt/vol) trichloroacetic acid. The reaction mixture was further incubated at 65°C for 15 min to hydrolyze N-formylkynurenine (N-FKyn) to l-kynurenine (l-Kyn). After centrifugation at 11,500 × g for 7 min, the clear supernatant (125 μl) was transferred into another 96-well plate and mixed with an equal volume of 2% (wt/vol) p-dimethylaminobenzaldehyde (DMAB) in acetic acid. The absorbance at 480 nm of the yellow pigment derived from the reaction with Kyn was measured using a Tecan multiplate reader (infinite M200; Tecan, Austria). A standard curve for concentrations of l-Kyn ranging from 0 to 500 μM was employed. The K_m of mIDO-2 was determined by measuring the activity at eight concentrations of l-Trp (0 to 30 mM). The absorbance of l-Trp was subtracted from the values. Each measurement was conducted in triplicate. For a direct comparison of the kinetic properties of the mIDO-2 WT and mutants, the assay was performed on the same 96-well plate. Kinetic parameters (K_m, maximum rate of metabolism [V_max]) were obtained using GraphPad Prism (version 4) software (GraphPad Software Inc., La Jolla, CA, USA).

Comparison of the K_m and V_max values for the mIDO-2 WT and mutant forms was performed using the unpaired t test or, for multiple comparisons, using a one-way analysis of variance statistical analysis followed by Tukey's honestly significant difference test and Games-Howell post hoc analysis, respectively. P values deemed significant are described in the table legend.

Screening of a natural compound library against mIDO-1.

The assay system described above was employed to screen a natural compound library (Leibniz Institute for Natural Product Research and Infection Biology, Jena, Germany) consisting of 2,000 compounds with activities against mIDO-1. The screening assay was performed in 96-well plates (Greiner) as described above, and the assay mixture contained 100 mM potassium phosphate buffer, pH 7.4, 100 mM KCl, 20 mM AA (neutralized with NaOH), 200 μg/ml catalase, 10 μM methylene blue, 400 μM l-Trp, and 75 μg/ml mIDO-1. The compounds were added to a final concentration of 5 μg/ml (∼10 μM). On each plate, the concentrations of kynurenine, positive controls without substrate, and negative controls without compounds used to create the reference values were included. For compounds showing >80% inhibition of the enzymatic activity, the 50% inhibitory concentration (IC₅₀) was determined in 96-well plates.

Synthesis of quinolizinium derivatives.

The quinolizinium derivatives consisting of compounds 2c and 2d (39), 2e and 3a to 3h (40), 4 (41), 2j (42), 2f and 2h (43), 2j (44), 5 (45), 2b (46), 2a (47), and 2g (48) were prepared according to published procedures (Fig. 1). Except for compound 3d, all compounds are sufficiently soluble in water (10⁻⁴ to 10⁻² M⁻¹). The log of the distribution coefficient (logD) values were determined for compound 2c (logD = −0.94) and compound 3e (logD = +0.20) as representative examples according to a published protocol (49).

Determination of IC₅₀s on mIDO-1 and -2.

For inhibitor testing, the IDO assay was performed as outlined above with slight modifications. The quinolizinium derivatives were dissolved in dimethyl sulfoxide (DMSO) and further diluted in the corresponding assay buffer for mIDO-1 (50 mM potassium phosphate, 100 mM KCl, pH 6.5) or mIDO-2 (100 mM potassium phosphate, 100 mM KCl, pH 7.4). A 2-fold serial dilution of the starting concentration of each inhibitor (from approximately 10 nM to 130 μM) was carried out in 96-well plates. l-Trp was added at 400 μM for mIDO-1 (2.5 μg/ml) and 4 mM for mIDO-2 (100 μg/ml). Due to the photosensitivity of the quinolizinium derivatives, the assay was performed under yellow light conditions. The tested quinolizinium derivatives are largely colored and fluorescent compounds in liquid solutions that absorb at 480 nm. Therefore, a negative control lacking the enzyme was measured under the same assay conditions for each compound. IC₅₀s were calculated using GraphPad Prism (version 4) software.

Homology modeling of mIDO and protein-ligand docking of benzoquinolizinium compounds.

The structures of mIDO-1 and mIDO-2 were predicted using the homology modeling tools of SWISS-MODEL (50, 51), including the Swiss-Pdb viewer for manual manipulation (15). The structures of both mIDO proteins were modeled on the basis of the crystal structure of hIDO-1 (PDB accession number 2D0T) (52). A comparison of the active-site structures of hIDO-1 and mIDO-1 revealed no prominent differences; wherefore, the coordinates of the heme group of hIDO-1 were transferred to mIDO-1 and employed for docking experiments.

Three-dimensional structures of the quinolizinium derivatives were generated with the SYBYL program package. The ligands were minimized by 1,000 cycles into a low-energy starting conformation using the SYBYL force field function of the Tripos program (version 7.2), and the results were saved as mol2 files. All hydrogen atoms in ligand and protein molecules had to be present for docking analyses. The protonation state was fixed to be protonated in Arg and Lys residues and deprotonated in Glu and Asp residues. The atom types for the ligands were assigned on the basis of the SYBYL function. The GOLD program (version 4.1) was used for docking studies, and scoring of all obtained binding geometries was carried out with the GOLD fitness functions used in different combinations. The binding site of the protein was defined to be 15 Å around the iron atom of the heme group via the implemented Ligsite algorithm (53). Furthermore, some of the geometries received from the docking runs were statistically analyzed with the DrugScore program (54), which uses about 30,000 crystal structures of the Cambridge Structural Database (http://www.ccdc.cam.ac.uk) as the data source. All results were visualized with the PyMOL system (55).

Cultivation of human tumor cell lines and determination of GI₅₀ values.

To assess cell growth inhibition by the quinolizinium derivatives, Jurkat (human T-cell leukemia) and A549 (small cell lung carcinoma) cells were tested as described in reference 56. These two cell lines are commonly used in the evaluation of potentially cytotoxic compounds. Moreover, from a physical point of view, Jurkat cells grow in a suspension, and A549 cells grow as an adhesion. These different forms of cell growth allow a more accurate profile of the cytotoxicity for different tissues to be obtained. Jurkat cells were grown in RPMI 1640 medium, and A549 cells were grown in Dulbecco modified Eagle medium supplemented with penicillin G (115 U/ml), streptomycin (115 μg/ml), and 10% fetal bovine serum. Individual wells of a 96-well tissue culture microtiter plate were inoculated with 100 μl complete medium containing 5 × 10³ cells. The plates were incubated at 37°C in a humidified incubator for 18 h prior to the experiments. After removal of the medium, 100 μl drug solution of various concentrations that had been dissolved in DMSO and diluted with complete medium was added, and the mixture was incubated at 37°C for 72 h. Cell viability was determined spectrophotometrically at a λ value of 570 nm (background correction at 630 nm) by the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) test as described previously (57). The ligands did not interfere with the photometric analysis of the MTT test, as they do not absorb significantly at a λ value of >550 nm.

The GI₅₀ was defined as the compound concentration required for inhibiting cell proliferation by 50% in comparison to the growth of cells treated with the maximum amount of DMSO (0.25%, vol/vol), which were considered to be 100% viable.

Cultivation of Plasmodium falciparum and determination of IC₅₀s.

Chloroquine (CQ)-sensitive strains of P. falciparum (3D7-Netherlands) were grown in continuous culture as described by Trager and Jensen (58) with slight modifications. Unless otherwise stated, parasites were maintained at 1 to 10% parasitemia and 3.3% hematocrit in RPMI 1640 culture medium (Gibco, Paisley, United Kingdom) supplemented with type A-positive erythrocytes, 0.5% lipid-rich bovine serum albumin (Albumax), 9 mM (0.16%) glucose, 0.2 mM hypoxanthine, 2.1 mM l-glutamine, and 22 μg/ml gentamicin. All incubations were performed at 37°C in 3% O₂, 3% CO₂, and 94% N₂. The synchronization of parasites at the ring stage was carried out via treatment with 5% (wt/vol) sorbitol (59).

Isotopic drug sensitivity assays by means of the semiautomated microdilution technique (60) were employed to investigate the susceptibility of the malaria parasites to various compounds. Radioactive [³H]hypoxanthine is taken up by the parasite as a precursor of purine deoxynucleotides for DNA synthesis (61). In a 96-well plate, a 2-fold serial dilution of each drug to be tested was carried out. Parasites were incubated at a parasitemia of 0.25% (>70% at ring stage) and 1.25% hematocrit in hypoxanthine-free medium. After 48 h, 0.5 μCi [³H]hypoxanthine was added, and the plates were incubated for another 24 h. Cells were harvested on a glass fiber filter (Perkin-Elmer, Rodgau-Jügesheim, Germany), washed, and dried. Their radioactivity (in counts per minute) is proportional to the growth of P. falciparum.

Murine Plasmodium berghei model.

The in vivo experiments described here were carried out at the Swiss Tropical Institute (Basel, Switzerland), adhering to local and national regulations of laboratory animal welfare in Switzerland. The study was approved by the Institutional Animal Care and Use Committee. The compounds were tested in the murine (NMRI mouse) P. berghei model essentially as described previously (62–64). The infection was initiated at day 0 with the P. berghei GFP ANKA malaria strain (a donation from A. P. Waters and C. J. Janse, Leiden University). Heparinized blood was taken from donor mice with approximately 30% parasitemia and diluted in physiological saline to 10⁸ parasitized erythrocytes/ml. An aliquot (0.2 ml) of this suspension was injected intravenously into experimental and control groups of mice. In untreated control mice, the level of parasitemia regularly usually rose to ∼30% by day 3 postinfection. Control mice died between day 6 and day 7 postinfection. In the experiments described herein, however, control animals (n = 5) were euthanized on day 4 postinfection for ethical reasons. Each of the quinolizinium derivatives tested (compounds 2c, 3f, 3g, and 3e) was applied orally (p.o.) and subcutaneously (s.c.) (at 4, 24, 48, and 72 h postinfection) to mice (n = 3) at a dose of 100 mg/kg of body weight per day. Parasitemia was determined on day 4 (96 h postinfection) and compared to that in control mice. Mouse survival was monitored every day for up to 30 days.

RESULTS

Identification of inhibitors of IDO-1.

In order to identify potential inhibitors of mIDO-1, we screened a natural compound library kindly provided by the Leibniz Institute for Natural Product Research and Infection Biology (Hans-Knöll Institute), Jena, Germany. The library consists of about 2,000 diverse compounds mainly derived from natural sources, and the activities of the compounds in the library against recombinant mIDO-1 were tested in a medium-throughput screening format using a spectrophotometric assay. The 8-aminobenzo[b]quinolizinium bromide (compound 2c) was identified to be the most potent inhibitor of mIDO-1 from this library, inhibiting mIDO-1 with an IC₅₀ of 160 nM. The IC₅₀s of two compounds (compounds 3a and 3b) could not be precisely determined, as the compounds precipitated in the assay at higher concentrations.

Subsequently, we synthesized and tested 21 quinolizinium derivatives of compound 2c (compounds 2 to 5). The derivatives could be classified into three main groups: (i) benzo[b]quinolizinium derivatives 2a to 2k with different types of substituents, namely, amino functionalities (compounds 2a to d, 2g, and 2k), a carboxylic acid (compounds 2f and 2h), a hydroxy group (compound 2i), a sulfanyl substituent (compound 2e), or a bromine atom (compound 2j); (ii) a 9-arylamino-substituted benzo[b]quinolizinium with a fixed quinolizinium unit and various substitution patterns at the phenyl ring; and (iii) quinolizinium derivatives with an additional heterocyclic fragment such as an annelated indole unit (compound 4) or a benzoimidazolyl substituent (compound 5). (Fig. 1 and Table 1). Determination of the IC₅₀s of the benzo[b]quinolizinium derivatives for mIDO-1 and mIDO-2 was performed using the optimized MB-AA assay under yellow light conditions. The initial compound, compound 2c, contains an amino group at the C-8 atom of the benzo[b]quinolizinium parent structure and with an IC₅₀ of 164 nM was the most potent inhibitor of mIDO-1 among the derivatives studied here. The N-aryl-9-aminobenzo[b]quinolizinium derivatives 3e, 3f, and 3g as well as the 9-methylsulfanylbenzo[b]quinolizinium derivative 2e also inhibit mIDO-1 at nanomolar concentrations (IC₅₀ range, 506 to 624 nM; Table 1). Derivatives with an amino residue at C-11 (compound 2g) or C-6 (compound 2a) showed only weak inhibitory properties for mIDO-1 (IC₅₀s, >128 μM). All other examined quinolizinium derivatives inhibited mIDO-1 at micromolar concentrations. In contrast, no relevant inhibitory effects of the compounds on mIDO-2 were observed at concentrations ranging from 10 nM up to 130 μM. Figure 2 shows an overlay of the crystal structure of human IDO-1 and the model of mIDO-2, which is discussed below with respect to inhibitor binding.

TABLE 1.

Structures and IC₅₀s of selected quinolizinium compounds for mIDO-1 and P. falciparum strain 3D7, as well as GI₅₀s for human tumor cell lines^a

^aThe IC₅₀s for recombinant mIDO-1 were determined using the MB-AA assay. The growth of P. falciparum strain 3D7 was determined by the [³H]hypoxanthine assay. IC₅₀s are in micromolar for mIDO-1 and in nanomolar for P. falciparum. Results are mean values ± standard deviations. Data were obtained from at least three independent experiments (n ≥ 3). OMe, methoxy; Me, methyl.

FIG 2.

Overlay of the crystal structure of hIDO-1 (teal) and the model of mIDO-2 (magenta). The active-site residues of mIDO-2 (Tyr231, Ala384, and Met385) that were mutated into the corresponding amino acid residues of hIDO-1 are highlighted.

Predicted modes of benzoquinolizinium compound binding to IDO-1.

By employing a homology model of mIDO-1, docking analyses of quinolizinium compounds with strong (compounds 2c and 3f), moderate (compound 3b), or low (compounds 2a and 2i) inhibitory properties were conducted by using the program GOLD. Thereby, the putative binding modes of the compounds in the heme-binding pocket of mIDO-1 were analyzed. Due to the predominantly hydrophobic binding pockets of heme-containing proteins or the direct interactions of ligands and the iron atom of heme groups, scoring with commonly used fitness functions can be only partially employed. In order to dock quinolizinium compounds into mIDO-1, two different fitness functions of GOLD (the original GoldScore and ChemScore functions), including parameter files improved previously (65), were used. For both fitness functions, notable differences in the positioning of ligands within the active site of mIDO-1 were observed. As previously shown for members of the P450 family, the binding pocket of mIDO-1 is nearly 100% hydrophobic. However, the docked compounds were rather hydrophilic. The absence of H donors or acceptors led to difficulties in scoring the position of the benzo[b]quinolizinium part of the ligand. Therefore, we mainly concentrated on external and internal van der Waals energies that have been taken into account for scoring. The three-ring system was docked into the active site statistically significantly more often, whereas the various moieties at the C-9 position were turned toward the solvent. Whether the nitrogen atom on position 5 of the benzo[b]quinolizinium structure contributes to binding and inhibition could not be shown. By scoring the interactions of ligands with the heme group, the central ferrum atom was disregarded independently of the fitness function used. Although the scoring of the geometries is based on a range of criteria, putative hydrogen bonds are mainly involved in this process. Figures 3A and B show a possible orientation of compound 2c in the binding pocket of mIDO-1. Altogether, the docked ligands fit well into the defined binding pocket of mIDO-1. Differences in mIDO-1 inhibition by ligands with highly similar structures, such as compounds 3b and 3f, can be explained not only by their in silico binding mode but also, perhaps, by their availability: compounds 3f and 3g are modified with halogen atoms in the para position of the phenyl ring, which forces them to slide into the binding pocket of mIDO-1 by entropic forces. In contrast, compound 3b contains a methoxy functionality on this position of the phenyl ring and is therefore able to form hydrogen bonds with surrounding water molecules and is kept in solution.

FIG 3.

Predicted modes of compound 2c binding to mIDO-1. (A and B) Predicted binding modes obtained using GoldScore (A) and ChemScore (B) as the fitness functions. (C) Best-fitting position of compound 3f.

Cytotoxicity tests on human cell lines in vitro.

The cytotoxicity of 8-aminobenzo[b]quinolizinium (compound 2c) and selected N-aryl-9-aminobenzo[b]quinolizinium derivatives (compounds 3a, 3b, 3c, 3e, 3f, 3g, and 3 h) was tested in vitro on two human tumor cell lines, namely, the Jurkat and A549 cell lines, revealing GI₅₀ values ranging from 1 to 28 μM (Table 1). Compound 2c showed the lowest cytotoxicity, with GI₅₀s of 62 μM and 86 μM for Jurkat and A549 cells, respectively. The effect of compounds 3e and 3g, with GI₅₀s ranging from 0.8 to 2.5 μM, also differed by 3 orders of magnitude from the 50% effective concentration (EC₅₀) for P. falciparum in cell culture. We selected the Jurkat T-cell and A549 cell lines since they are commonly used in the evaluation of potentially cytotoxic compounds. Moreover, from a physical point of view, Jurkat cells grow in a suspension and A549 cells grow as an adhesion. These different forms of cell growth allow a more accurate profile of the cytotoxicity for different tissues to be obtained.

Inhibition of P. falciparum growth in vitro.

The growth-inhibitory effects of quinolizinium derivatives 2 to 5 on CQ-sensitive P. falciparum strain 3D7 were determined in vitro (Table 1). The N-aryl-9-aminobenzo[b]quinolizinium derivatives 3b, 3c, 3e, 3f, 3g, and 3h showed the highest antiplasmodial activity, with IC₅₀s of <10 nM, and inhibited parasite growth more effectively than compound 2c (IC₅₀, 109 nM). All other derivatives required higher concentrations to achieve the same growth inhibition. The derivatives with the strongest antimalarial activity were quinolizinium salts containing bromide as the anion. Therefore, the effect of sodium bromide (NaBr) on the growth of P. falciparum was examined. However, at concentrations comparable to those of the quinolizinium derivatives, no inhibitory effect of the bromide on P. falciparum cell culture was observed (data not shown).

In vivo efficacy of quinolizinium derivatives.

Compound 2c and three quinolizinium derivatives with the strongest antiplasmodial activity and inhibitory effects on mIDO-1 in vitro (compounds 3e, 3f, and 3g) were tested in a murine P. berghei model. Each compound was administered to mice (n = 3) orally and subcutaneously (at 4, 24, 48, and 72 h postinfection) at a dose of 100 mg/kg per day. After oral treatment, none of the compounds showed significant antimalarial activity in vivo either by reducing parasitemia or by prolonging the survival of the animals. Subcutaneous application of the quinolizinium derivatives showed acute toxic effects.

Structural and functional differences of IDO-1 and -2.

To study the reasons for the different substrate and inhibitor specificities of IDO-1 and -2, an overlay of mIDO-2 with the crystal structure of hIDO-1 (PDB accession number 2D0T) was prepared as previously described (15). The overall structure is similar to that of hIDO-1, and the enzymes show a 43% amino acid identity (15). The secondary structures of mIDO-2 and hIDO-1 are similar but show several major differences in the active site and the entrance tunnel of the heme-binding pocket. The substrate-binding pocket of hIDO-1 is bigger than that of mIDO-2 (15), which might be responsible for their different substrate specificities and affinities as well as their different catalytic activities. An important difference in the heme environment is the replacement of Leu384 of hIDO-1 by Met385 in mIDO-2 at the entrance of the tunnel: the bulky Met385 most likely sterically impairs the access of substrates to the heme-binding site in mIDO-2 (Fig. 2). The residues Phe226, Phe227, and Arg231 are supposed to be critical for dioxygenase activity in hIDO-1 (52). Phe227 is a part of the entrance tunnel to the heme-binding pocket and was suggested to be involved in substrate recognition by hydrophobic interactions (Fig. 2) (15). Phe227 corresponds to Tyr231 in mIDO-2. Moreover, the negative charge of Asp383 of hIDO-1 (Ala384 in mIDO-2) appears to be involved in substrate recognition (Fig. 2).

To investigate the function of certain active-site residues, we generated three active-site mutants of mIDO-2 (mIDO-2^M385L, mIDO-2^Y231F, mIDO-2^A384D) using site-directed mutagenesis. The mutants were heterologously overexpressed and purified as described above for the mIDO-2 wild type. It had been shown previously that the MB-AA system is not the optimal reducing system for assaying IDO-2 activity (15). However, the system was chosen for use here in order to allow a comparison of IDO-2 to IDO-1 and the results of previous studies. Care was taken to use identical protein concentrations, allowing a direct comparison of the wild type and the mutants. Determination of the K_m and V_max of the mIDO-2 mutants for direct comparison to those of the wild type was also performed. Because mIDO-2 is unstable, only freshly purified protein was used at two different concentrations of 75 and 125 μg/ml. Under these conditions, mIDO-2 had a K_m of about 11 mM for l-Trp (Table 2), which is in the range of K_m values reported for hIDO-2 (66). Statistical analysis revealed significant differences between the K_m values for Trp of the mIDO-2 WT (11.2 mM), mIDO-2^M385L (6.48 mM), and mIDO-2^Y231F (15.6 mM) determined at a protein concentration of 125 μg/ml. The K_m of the mIDO-2 WT is in good accordance with previously reported data (15). At a final concentration of 75 μg/ml enzyme, the K_m as well as the V_max value of mIDO-2^M385L (7.41 mM and 53.7 mU/mg, respectively) were significantly lower than those of mIDO-2 WT (21.8 mM and 108 mU/mg, respectively). The K_m and k_cat/K_m of mIDO-2^M385L were higher than those of the mIDO-2 WT (Table 2). The K_m and V_max of mIDO-2^A384D and the mIDO-2 WT were not significantly different.

TABLE 2.

Kinetic parameters of the mIDO-2 WT and the mIDO-2^Y231F, mIDO-2^A384D, and mIDO-2^M385L active-site mutants^b

Strain	mIDO-2 concn (μg/ml)	Km (mM)	Vmax (mU/mg)	kcata (min−1)	kcat/Km (min−1 mM−1)
mIDO-2 WT	125	11.2 ± 1.90	54.9 ± 15.7	2.47 ± 0.71	0.22 ± 0.04
	75	21.8 ± 4.92	108 ± 17.8	4.85 ± 0.80	0.22 ± 0.05
mIDO-2Y231F	125	15.6 ± 2.66*	68.8 ± 13.2	3.10 ± 0.59	0.20 ± 0.02
	75	25.7 ± 0.11	121 ± 7.64	5.46 ± 0.34	0.21 ± 0.01
mIDO-2A384D	125	12.1 ± 2.61	59.2 ± 5.30	2.67 ± 0.24	0.22 ± 0.05
	75	27.2 ± 10.3	102 ± 38.3	4.61 ± 1.72	0.17 ± 0.02
mIDO-2M385L	125	6.48 ± 1.20**	42.3 ± 11.0	1.90 ± 0.49	0.29 ± 0.08
	75	7.41 ± 3.32**	53.7 ± 8.06**	2.42 ± 0.29	0.33 ± 0.21

^ak_cat (turnover number) is expressed in micromoles of product per micromole of holoenzyme per minute.

^bResults are expressed as mean values ± standard deviations and were obtained from at least three independent experiments. *, P < 0.05; **, P < 0.01.

DISCUSSION

The kynurenine pathway, in which IDO is one of the major enzymes, is highly interesting with respect to the pathogenesis of several diseases, including parasitic infections such as CM. We identified an 8-aminobenzo[b]quinolizinium bromide (compound 2c) and its derivatives to be potent inhibitors of both mIDO-1 and the human malaria parasite P. falciparum in culture and studied their effects on human tumor cell lines, on heme polymerization, and in a murine malaria model. Additionally, we studied the enzymatic mechanism of mIDO-2 with an in silico modeling approach, site-directed mutagenesis, and kinetic analyses in more detail.

Quinolizinium derivatives as potent inhibitors of mouse IDO.

A range of IDO-1 inhibitors has already been discovered in in vitro assay systems (16, 67–69) or a mouse tumor model (70). As reported by Meininger et al. (16), IC₅₀s for human IDO-1 found in a Bridge-IT tryptophan fluorescence assay were 35.6 μM for N-methyl-l-Trp, 21.6 μM for methyl thiohydantoin-Trp, and 3.0 μM for Amg-1, with Amg-1 showing a clear selectivity for IDO-1 over IDO-2 (16). Furthermore, a number of potent IDO inhibitors, including annulins A (K_i, 0.69 μM), B (K_i, 0.12 μM), and C (K_i, 0.14 μM), were identified in the marine hydroid Garveia annulata (67). Rohrig et al. (68) used computational structure-based methods to optimize the activity of the IDO-1 inhibitor 4-phenyl-1,2,3-triazole. This approach yielded a series of low-molecular-weight inhibitors, with the most active compound having potency at nanomolar concentrations. Furthermore, Carr et al. (69) prepared synthetic analogues of the sponge natural product exiguamine, with the best compounds being active against recombinant human IDO at concentrations between 40 and 400 nM.

In our study, the 8-aminobenzo[b]quinolizinium bromide (compound 2c) derived from an in vitro screening against a natural compound library was found to inhibit mIDO-1 at nanomolar concentrations (IC₅₀, 164 nM). None of the 21 derivatives synthesized and studied was more active against IDO-1 than the original compound 2c (Table 1), but interestingly, some compounds were much more potent in inhibiting the growth of P. falciparum. A main structural difference between compound 2c and the derivatives with the weakest inhibitory activity (compounds 2a, 2d, and 2g) is the position of the amino group at the benzo[b]quinolizinium parent structure: in compound 2c, the NH₂ group is located at C-8 of the ring system, while it is located at C-9 in compound 2d (IC₅₀, 18 μM), at C-6 in compound 2a (IC₅₀, >128 μM), and at C-11 in compound 2g (IC₅₀, >128 μM). The N-aryl-9-aminobenzo[b]quinolizinium derivatives 3e, 3f, and 3g and the 9-methylsulfanylbenzo[b]quinolizinium derivative 2e inhibited mIDO-1 at concentrations comparable to the inhibitory concentration of the original compound (IC₅₀ range, 506 to 624 nM), while all other derivatives showed potency at micromolar concentrations. Thus, the addition of an aromatic benzene ring bound to the amino group at C-9 or of different other groups to the basic ring system affects the inhibition of mIDO-1.

Additionally, we examined the binding mode of the quinolizinium compounds via in silico docking analyses and observed multiple binding modes for the ligands, but they had distinct preferences for the positions described above (Fig. 3A and B). Analyses of the binding of compound 3f, one of the most potent inhibitors, showed that aromatic amino acid residues, such as Phe230 or Phe167, and the aliphatic Val134, as well as the carboxylate group of the heme molecule, appear to be involved in inhibitor binding (Fig. 3C).

Potential of inhibition of IDO-1 in pharmacological treatments.

The kynurenine pathway and IDO-1 have been linked to the pathogenesis of CM in several studies of humans (23, 24, 71) and mice (9, 10, 25–29). Moreover, IDO-1 has been suggested to be a target for therapeutic interventions in other diseases characterized by pathological immune suppression. Although mice lacking IDO-1 were not protected against CM, partial protection was observed in C57BL/6 mice treated with an inhibitor of kynurenine-3-hydroxylase, the subsequent enzyme in the kynurenine pathway (9). The protection was associated with decreased levels of picolinic acid in the brain but not with changes in the levels of kynurenic acid or quinolinic acid. Although the behavioral changes, histopathology, and immunological manifestations seen in murine models of CM show similarities to the characteristics of human CM, there are differences, such as the cell types that adhere to the cerebral microvascular endothelium (29). Thus, the roles of IDO-1, IDO-2, and the metabolites of the kynurenine pathway in the pathogenesis of CM in humans need to be studied in further detail.

Interestingly, the IDO inhibitors identified in our screening approach were also highly active against P. falciparum 3D7 blood stages in cell culture (IC₅₀, <10 nM; Table 1). Red blood cell lysis was not observed at the inhibitor concentrations used. Pathogenic bacteria such as Pseudomonas aeruginosa and Burkholderia cepacia have IDO homologs (31). However, searches using BLAST analysis did not identify close homologs of IDO in Plasmodium, T. gondii, and L. major. Thus, an IDO homolog is unlikely to be the target of quinolizinium derivatives in P. falciparum. As described for structurally related compounds before (36), DNA binding represents a potential mechanism of action. With respect to the structure-activity relationship, a common feature of the quinolizinium derivatives most active against P. falciparum was the 9-arylamino substituent. However, the corresponding counteranion might also play a role, as this is the only difference between compound 3e (IC₅₀, 3.7 nM; Br⁻) and compound 3d (IC₅₀, 93.3 nM; BF₄⁻). In this particular case, the hydrophobic tetrafluoroborate anion in compound 3d may lead to the very low solubility of the compound in water, thus leading to precipitation or aggregation. Furthermore, the position of the NH₂ group in the parent ring structure appears to be important for the inhibition properties, with compound 2d, which has an NH₂ group at C-9, being more potent than compounds 2a and 2g. Hydrophilic functionalities, such as the hydroxyl group (compound 2i) and the carboxy substituent (compounds 2f and 2h), diminish the antimalarial potency.

The derivatives 2c, 3e, 3f, and 3g were, furthermore, tested for their antimalarial action in a murine P. berghei model. Compound 2c applied orally did not show an effect on parasite clearance or the life span of the infected mice, which was likely due to rapid renal elimination. The increased hydrophobicity obtained by adding an aromatic benzene ring to the basic ring structure, as in compounds 3e, 3f, and 3g, did not enhance the effects in the mouse model, which might be explained as follows: the compounds under investigation carry a permanent positive charge due to the quaternary bridgehead nitrogen atom and can be converted into a derivative with a neutral charge only by deprotonating an acidic functionality such as a carboxylic acid (compounds 2f and 2h) or a hydroxy substituent (compound 2i), and even in these cases, the conjugate bases are, rather, zwitterions. All other quinolizinium derivatives are generally not able to form a neutral tautomer or conjugate base under neutral physiological conditions. Those compounds will be unlikely to have passive permeation, and as such they have very low oral bioavailability. Given that poor permeation limits entry into a cell, the cytotoxicity data should be carefully interpreted. Hence, the logD data for representative derivatives and compound 3e (logD = +0.20) show that the parent amino-substituted benzo[b]quinolizinium, compound 2c (logD = −0.94), exhibits a hydrophilic character due to its permanent ionic nature, but notably, the introduction of an additional aryl ring in derivative 3e significantly increases its lipophilicity.

Furthermore, when the high efficiency of the compounds against asexual P. falciparum blood stages are compared to their low levels of activity in the P. berghei mouse model, differences between the two parasites need to be taken into account. P. falciparum induces in its host erythrocytes new permeability pathways (NPPs) (72), and permanently positively charged molecules can enter the infected erythrocyte and kill the parasite. P. berghei does not induce NPPs, and so compounds are active against P. berghei blood stages only if they can tautomerize to a neutral species. Furthermore, a compound targeting IDO in the brain would need to cross the blood-brain barrier. Notably, the cytotoxic effects of the quinolizinium derivatives on the mammalian cells occurred only at concentrations that were about 3 orders of magnitude higher than those required to inhibit parasite growth in vitro. Again, limited membrane permeability might contribute to this result. However, the fact that the compounds of the benzo[b]quinolizinium series have three planar aromatic rings and as such are similar to the structurally related acridines should also be taken into account. Acridines are DNA intercalators and therefore mutagenic (73), a characteristic that might contribute to the effects of the quinolizinium compounds on P. falciparum asexual blood stages. Nevertheless, we believe that it is worth reporting on and following up these analyses of this interesting class of compounds, which strongly inhibit P. falciparum growth and act as IDO inhibitors. Thus, in the next step we plan to further optimize the compounds, assess their permeation/efflux potential, look at brain/plasma binding and brain/plasma ratios, and carry out more detailed toxicological studies.

Catalytic mechanism of mammalian IDO-2.

Recently, different tryptophan analogues were tested as the substrates for human IDO-2, with 5-methoxy-tryptophan having the highest affinity (K_m, 547 μM), while l-Trp was the sixth most active substrate of the tested analogues (66). Our study confirms the results of previous studies reporting that IDO-2 is an enzyme with comparatively low activity that might be activated under currently unknown conditions, depending on cell types and regulatory molecules or modifications in vivo (66). In transiently transfected HEK193T cells supplied with tryptophan, IDO-2 had 50% of the activity of IDO-1 (14), demonstrating that the huge difference between the activities of the two isoforms measured in cell-free assays does not necessarily reflect their in vivo activities (15). Furthermore, it should be taken into account that MB-AA is not the ideal reducing system for the analysis of IDO-2 activity in vitro, as previously discussed (13–15, 74, 75). We chose to use the system nonetheless in order to be able to compare our data with those from previous analyses, but we believe that the differences in the K_m and V_max values at different enzyme concentrations (Table 2) might be due to the inefficient reduction of IDO-2 by MB. However, with one enzyme concentration, a direct comparison of the different mutants of mIDO-2 using the MB-AA-based assay is possible.

Interestingly, the quinolizinium compounds did not show major inhibitory effects on mIDO-2. High concentrations of enzyme and l-tryptophan as well as low enzymatic activity might contribute to this result. Additionally, a structural reason for the lack of inhibitory effects on mIDO-2 is likely to be the bulky Met385 of mIDO-2 at the entrance of the pore, which prevents the relatively large and sterically demanding quinolizinium derivatives from entering the active site of the enzyme.

In order to further analyze the catalytic mechanism of mammalian IDO-1 and IDO-2, we created a homology model of mIDO-2 based on the crystal structure of hIDO-1 (52). Based on this, we created enzyme mutants using site-directed mutagenesis and inserted the corresponding amino acids of IDO-1 into the active site (mIDO-2^Y231F, mIDO-2^A384D) or the entrance tunnel to the active site (mIDO-2^M385L) to analyze differences in substrate binding between IDO-1 and IDO-2. Phe227 from hIDO-1 is replaced by Tyr231 in mIDO-2 (Fig. 2), which was suggested to make the heme-binding pocket more susceptible to phosphorylation or sulfation (15). mIDO-2^Y231F has a slightly decreased affinity to l-Trp but did not show any significant change in specific activity compared to that of the wild type (Table 2), indicating that Tyr231 is involved in but not crucial to substrate binding. Moreover, Phe is not more effective than Tyr in this position. The polar Asp383 of hIDO-1 is replaced by the apolar aliphatic Ala384 in mIDO-2, with the negative charge of Asp potentially being involved in substrate recognition (Fig. 2). A mutation of Ala384 of mIDO-2 into an Asp has, however, no influence on substrate affinity or activity (Table 2).

The entrance tunnel leading to the heme-binding site is narrower in mIDO-2 than in hIDO-1, since the Met at position 385 seems to limit the access of the substrates in the active site. Indeed, a mutation of Met385 in mIDO-2 led to a significant increase in affinity toward l-Trp (Table 2). This is most likely due to an improved accessibility of the active site. The narrow tunnel to the heme-binding site might be indicative of a preference of IDO-2 for a currently unknown substrate smaller than l-Trp. Accordingly, the observation that the mutation of Tyr231 and Ala384 into the corresponding residues of hIDO-1 shows no or only minor influences on the l-Trp-degrading activity of mIDO-2 points toward the distinct substrate preference not being due to differences in single active-site amino acids. The facts that IDO-1 and IDO-2 show different expression patterns, are differentially expressed during malaria infections, and have distinct catalytic properties suggest that they can be selectively targeted by inhibitors (13, 14, 74). Since the physiological function of IDO-2 is not yet fully known, selective IDO-2 inhibitors will be highly useful in determining its in vivo function (15).

Conclusion.

In summary, we identified N-aryl-9-aminobenzo[b]quinolizinium derivatives that (i) are effective inhibitors of mammalian IDO-1 and (ii) are exceptionally active against P. falciparum in cell culture. Although we do not yet know their exact mechanism of inhibition toward parasites, an inhibitor that affects the growth of P. falciparum directly and at the same time inhibits IDO-1 from the host in order to reduce pathophysiological effects and symptoms might be a highly beneficial combination for treating patients suffering from cerebral malaria. Further studies will focus on further optimizing the IDO-1-inhibitory action of the compounds, on improving their bioavailability, and on systematically assessing the origin of their toxicity. Assessment of the origin of their toxicity is necessary to find ways to decrease their cytotoxic properties. At the same time, the bioavailability of the compounds may be increased by introducing appropriate functional groups (76, 77) or by using host guest-based drug delivery systems such as lipids, cyclodextrins, or liposomes (78, 79).